Children exposed in the womb were no more likely to be diagnosed with autism or ADHD than their own brothers and sisters. But the same records showed a raised risk among mothers who took paracetamol before they conceived — pills that never reached the baby. That second result is the useful one.
Paracetamol is the painkiller and fever treatment recommended in pregnancy across the UK, and the regulator restated that after claims made in the United States in September 2025.4 The awkward fact behind the argument is that women who take painkillers are not like women who don't — they may be more unwell, more stressed, or carrying conditions that run in the family. Any of that could explain a difference in their children that the drug did not cause.
A study published in JAMA Internal Medicine on 29 June 2026 got around this.1 From 708,020 Hong Kong mother–child pairs recorded between 2001 and 2023, the researchers picked out families where a mother took paracetamol in one pregnancy but not another, and compared those siblings with each other. Brothers and sisters share their genes, their mother and their home, so whatever else differs between families is held still.
The result is reported as a risk ratio, where 1.00 means the exposed and unexposed children were diagnosed at exactly the same rate. Across 124,333 children followed for autism and 97,285 for ADHD, the figures were 1.00 and 1.01 — no difference.
Then the researchers ran a test designed to fail. They repeated the whole analysis on paracetamol dispensed before the pregnancy started — a pill swallowed before there was a baby to affect. If the method were clean, that should have come out at 1.00 too. It came out at 1.12 for autism and 1.24 for ADHD.1
Hong Kong is not an outlier. Sweden ran the same sibling comparison across 2,480,797 children in 2024 and got 0.98 for autism.2 Taiwan did it across 2,092,926 births in 2026 and also got 0.98.37 A review pulling the international evidence together, published in The Lancet Obstetrics, Gynaecology & Women's Health in January 2026, found no link to autism, ADHD or intellectual disability.5
One caution on the headline number. Hong Kong's 1.00 looks like the cleanest of the three but is the least certain: its plausible range runs from 0.91 to 1.11, almost twice as wide as Sweden's, because far fewer families qualified. Landing on a round number is not the same as measuring precisely.
Nothing has changed for anyone who is pregnant. The MHRA's position is that paracetamol remains the first-choice painkiller in pregnancy, taken at the lowest dose for the shortest time, and that patients should not switch to ibuprofen, which is generally not recommended.4 It also makes the point that tends to go missing: an untreated fever carries risks of its own for the baby, so declining treatment is not the cautious, neutral option it can feel like.4
All three studies count prescriptions, not tablets actually swallowed, so paracetamol bought over the counter is invisible and real exposure is undercounted. Diagnoses come from health registries, so a child never assessed is counted as unaffected. And none of this covers paracetamol given to babies after birth: that has been studied mainly in hospital groups of a few hundred to a few thousand premature infants, nothing close to this scale or design.9
Comparing siblings is the best available way to strip out family effects, and the association vanishes in every large study that has done it. The Hong Kong records also show the machinery of the false result directly.12
The Taiwanese team found their own sibling results flipped depending on birth order: a raised risk when only the older child was exposed, a lowered one when only the younger was. They read that as a sign the method has problems of its own, and say it stops firm conclusions being drawn from it.6